Smoothest paths for different settings of EventStudy

We construct event studies using different numbers of leads and lags. We solve for the smoothest path in each case.

Summary data

A summary csv can be found at data_paths.csv. Recall that the order of the computed polynomial equals order+1.

  • The solver converges in all 72 cases.
  • A computation is successful if the absolute difference between the Wald critical value and the Wald optimal value is less than 0.1. There were 41 successes.

Order found and number of successes, by outcome:

knitr::kable(
  dt[, .(Num_found = .N,
         Num_success = sum(success)),
     by = .(yvar, order)]
)
yvar order Num_found Num_success
y_smooth_m 1 7 7
y_smooth_m 2 18 3
y_smooth_m 3 10 10
y_smooth_m 4 1 1
y_jump_m 4 3 3
y_jump_m 5 10 7
y_jump_m 6 16 8
y_jump_m 7 6 1
y_jump_m 8 1 1

Smooth outcome

Post=2

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7

Post=3

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7

Post=4

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7

Post=5

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7

Post=6

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7

Post=7

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7

Jump outcome

Post=2

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7

Post=3

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7

Post=4

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7

Post=5

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7

Post=6

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7

Post=7

Pre=2 Pre=3 Pre=4
Pre=5 Pre=6 Pre=7